Combining local and non-local shape descriptor for multi-modal image registration

Registration of multi-modal image pair provides complementary information for further analysis. Previous registration methods borrowed from mono-modal registration methods focused on local descriptors, making them inappropriate for multi-modal case. This paper proposes a novel image registration which combines local information and non-local information as well. Through this combination, more image structure information is used for multi-modal image registration. Experiments on real optical-SAR image pair demonstrated our method´s feasibility.